-
Glasgow Theses Service http://theses.gla.ac.uk/
[email protected]
n
Cross, Adam D. P. (2014) The influence of seabird-derived
nutrients on island food-webs. PhD thesis.
http://theses.gla.ac.uk/6312/ Copyright and moral rights for this
thesis are retained by the author A copy can be downloaded for
personal non-commercial research or study, without prior permission
or charge This thesis cannot be reproduced or quoted extensively
from without first obtaining permission in writing from the Author
The content must not be changed in any way or sold commercially in
any format or medium without the formal permission of the Author
When referring to this work, full bibliographic details including
the author, title, awarding institution and date of the thesis must
be given
-
The influence of seabird-derived nutrients
on island food-webs
Adam D. P. Cross
Submitted in fulfilment of the requirements for the
Degree of Doctor of Philosophy
Institute of Biodiversity, Animal Health and Comparative
Medicine,
College of Medical, Veterinary and Life Sciences
University of Glasgow
September 2014
-
Declaration
I declare that the work described in this thesis is of my own
composition and has been
carried out entirely by myself unless otherwise cited or
acknowledged. No part of this
thesis has been submitted for any other degree or
qualification.
The following chapter has been published in co-authorship with
my supervisor and others:
Chapter 6: published as Cross, A. D. P., Hentati-Sundberg, J.,
Osterblom, H., McGill, R.
A. R. and Furness, R. W., 2014. Isotopic analysis of island
House Martins Delichon urbica
indicates marine provenance of nutrients. Ibis, 156, 676 –
681.
Adam Cross
Glasgow 2014
-
2
Abstract
There is an increasing understanding of the influence seabirds
have on island food webs
globally, which often arises from the trans-boundary input of
nutrients. Seabird-derived
nutrients, primarily in the form of guano, can have significant
effects on island
communities by increasing primary productivity and then
indirectly influencing other
species. However, there are few studies looking at how the
influence of seabirds permeates
island food webs to higher trophic levels, in particular within
the United Kingdom, which
holds globally significant populations of seabirds. To
understand the extent to which
seabirds influence islands, the size of the seabird population
must be first reliably
determined. With an increasing seabird population size and
density the effects of seabirds
on land increases concomitantly.
The Atlantic puffin Fratercula arctica is a difficult species to
reliably monitor,
given its underground presence from the use of burrows and its
notoriously erratic
attendance at colonies. This study looks firstly at a novel
method to monitor the Atlantic
puffin using time-lapse photography. Time-lapse photography
provides a way to derive an
estimate of population size from counts of individuals, by
repeated photographs across a
period of time with relatively low cost and from areas normally
considered inaccessible.
The results showed there was a significant and positive
relationship between the maximum
numbers of Atlantic puffins observed and the size of the
population; further work is
required though to reduce the error associated with population
size estimates. Data from
high temporal resolution time-lapse photography shows how the
attendance of Atlantic
puffins at the colony varies over different temporal scales.
Given the variability in
sampling intensity the study stresses the need for standardised
sampling intensity with the
use of photography to monitor Atlantic puffins.
Secondly, this study showed how the presence of two seabird
species, the Atlantic
puffin and the great skua Stercorarius skua, alters island food
webs. These seabird species
are likely to change plant community diversity, relative to
areas without seabirds. The
chemical concentration of grasses inside seabird colonies was
also altered: grasses had
significantly higher concentrations of nitrogen and also had
higher values of δ15
N, relative
to areas without seabirds. These chemical alterations suggest
that nutrients from seabirds
are incorporated into local vegetation. Furthermore, samples of
hair from rabbits and sheep
found within puffin colonies also had significantly higher
values of δ15
N, suggesting that
nutrients travel from seabirds into secondary consumers, via
ornithogenic forage. An
-
3
additional study on the transfer of nutrients within island food
webs showed how
ornithogenic nutrients deposited on an island in the Baltic Sea
were incorporated into
house martins, via aquatic insects.
These studies, along with an understanding of seabird population
size, suggest that
the impact of seabirds on island food webs may be considerable
and have large
consequences for island conservation and management.
-
4
Table of Contents
Acknowledgements
................................................................................................................
8
Introduction
..........................................................................................................................
10
Chapter 2 - Variation in Atlantic puffin Fratercula arctica
colony attendance; monitoring
using time-lapse photography.
....................................................................................
23
Introduction
...........................................................................................................
25 2.1
Methods
.................................................................................................................
28 2.2
Results
...................................................................................................................
38 2.3
Discussion
.............................................................................................................
54 2.4
Chapter 3 - Monitoring Atlantic puffins Fratercula arctica using
time-lapse photography . 62
Introduction
...........................................................................................................
64 3.1
Methods
.................................................................................................................
67 3.2
Results
...................................................................................................................
73 3.3
Discussion
.............................................................................................................
84 3.4
Chapter 4 - A bioenergetic model estimating the nutrient input
from Atlantic puffins
Fratercula arctica and great skuas Stercorarius skua onto
Mingulay ........................ 93
Introduction
...........................................................................................................
95 4.1
Methods
.................................................................................................................
98 4.2
Results
.................................................................................................................
107 4.3
Discussion
...........................................................................................................
112 4.4
Chapter 5 - The influence of Atlantic puffins Fratercula arctica
and great skuas
Stercorarius skua on Scottish islands
.......................................................................
123
Introduction
.........................................................................................................
125 5.1
Methods
...............................................................................................................
129 5.2
Results
.................................................................................................................
142 5.3
Discussion
...........................................................................................................
154 5.4
Chapter 6 - Isotopic analysis of island House Martins Delichon
urbica indicates marine
provenance of
nutrients.............................................................................................
163
Introduction
.........................................................................................................
165 6.1
Methods
...............................................................................................................
167 6.2
-
5
Results
.................................................................................................................
169 6.3
Discussion
...........................................................................................................
172 6.4
Chapter 7 – General Discussion...
......................................................................................
176
APPENDIX I
......................................................................................................................
181
APPENDIX II
....................................................................................................................
185
APPENDIX III
...................................................................................................................
188
List of References
..............................................................................................................
195
-
6
List of Tables
Table 2.1. Summary statistics of the photographs and number of
puffins in photographs for
each
colony...........................................................................................................................
43
Table 2.2. Model coefficients for the analysis of the effect of
breeding stage, daily period
and colony on attendance of Atlantic puffins.
.....................................................................
44
Table 2.3. Model coefficients for the analysis of the effect of
daily period and colony on
attendance of Atlantic puffins during the pre-hatching stage.
............................................. 45
Table 2.4. Model coefficients for the analysis of the effect of
daily period and colony on
attendance of Atlantic puffins during the post-hatching stage.
............................................ 45
Table 2.5. The change in attendance in response to environmental
variables, before and
during an above-average
event.............................................................................................
46
Table 3.1. The area, density of apparently occupied burrows
(AOBs) and the ratio of AOBs
to the maximum number of puffins seen (Pmax) pre-hatch and
post-hatch ........................ 76
Table 3.2. Results of linear regression and its predictive
ability for the relationship between
puffin attendance and the number of Apparently Occupied Burrows
(AOBs). ................... 77
Table 4.1. Bioenergetics model parameters for the Atlantic
puffin and great skua. .......... 105
Table 4.2. Parameter definitions for bioenergetics model.
................................................ 106
Table 4.3. The nitrogen and phosphorous input per individual and
population per year on
the colony for the Atlantic puffin and great skua.
..............................................................
109
Table 4.4. Sensitivity analysis for the bioenergetics model for
the Atlantic puffin. .......... 110
Table 4.5. Sensitivity analysis for the bioenergetics model for
the great skua. ................. 111
Table 5.1. Details of study islands.
....................................................................................
141
Table 5.2. Mean richness, diversity and evenness indices for
plant data collected over two
successive years at different locations on Mingulay.
......................................................... 148
Table 5.3. The plant community mean percentage similarity
between sites. .................... 148
Table 5.4. Model coefficients for the nitrogen content of
Composite and Holcus
samples.
..............................................................................................................................
149
Table 5.5. Model coefficients for the nitrogen stable isotope
ratio of Composite and Holcus
samples.
..............................................................................................................................
149
Table 5.6. The mean nitrogen concentration and δ 15
N of Holcus lanatus and composite
samples across the three study islands.
..............................................................................
150
Table 5.7. The change in vegetation growth within and outwith
(open) exclosures over
time.
....................................................................................................................................
151
Table 5.8. Model coefficients for the change in vegetation
height with the effect of
exclosures
...........................................................................................................................
151
Table 6.1. The difference in isotope values between pairs of
species. .............................. 170
file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044563file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044563file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044569file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044569file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044573file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044574file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044575file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044580file:///C:/Users/Adam/Dropbox/Thesis/Final%20Drafts/With%20corrections/MS10_Final_draft_corrections7.docx%23_Toc417044580
-
7
List of Figures
Figure 2.1. Location of the three islands where cameras were
installed. Top left: UK
overview of island
locations.................................................................................................
37
Figure 2.2. Photographs of each colony.
..............................................................................
47
Figure 2.3. The change in attendance of Atlantic Puffins
throughout the day with data
across the study duration averaged into half-hourly intervals
(line) and into five daily ~5-
hour periods (bars)..
.............................................................................................................
49
Figure 2.4. The half-hourly (line) and ~5-hourly periods of
daily attendance (bars) of
Atlantic Puffins across the pre-hatch (left) and post-hatch
(right) periods .......................... 50
Figure 2.5. The mean change in the half-hourly mean numbers of
Atlantic puffins from the
pre- and post-hatching stages over colonies on Mingulay in 2011
and 2012. ..................... 51
Figure 2.6. The seasonal change in attendance of Atlantic
puffins across colonies. ........... 52
Figure 2.7. The scaled, de-trended attendance pattern of
Atlantic puffins over three
colonies on Mingulay in 2012.
.............................................................................................
53
Figure 3.1. A photograph of the Original colony in 2011 showing
the overlaid grid to mark
the position of Atlantic puffins.
...........................................................................................
78
Figure 3.2. The change in the maximum number of Atlantic puffins
recorded over the
whole season with varying photographic interval.
...............................................................
79
Figure 3.3. The change in daily maximum number of Atlantic
puffins with sampling
duration using a photographic interval of 40 min.
...............................................................
80
Figure 3.4. The results of linear regression between: A) Pmax
and B) Pmean (using a 40 min
interval) against the number of apparently occupied burrows.
............................................ 81
Figure 3.5. Kernel density distribution of Atlantic puffins
during the incubation period on
land in relation to apparently occupied burrows (AOB) on the
colony termed Original on
Mingulay 2011.
....................................................................................................................
82
Figure 3.6. The probability density function of the distance of
Atlantic puffins to their
nearest burrow.
.....................................................................................................................
83
Figure 5.1. Location of the three islands where studies were
conducted.. ......................... 140
Figure 5.2. NMDS plot of community plant data between sites in
2012 produced using a
Bray-Curtis similarity matrix.
............................................................................................
152
Figure 5.3. Mean values of δ15
N of grass samples and hair samples from within and
outwith seabird colonies on islands over time..
.................................................................
153
Figure 6.1. Location of the sampling sites in the Baltic Sea.
............................................. 171
Figure 6.2. Stable isotope values in solid black symbols for
bird feathers.. ...................... 171
-
8
Acknowledgements
I must first thank my supervisors, Dr. Ruedi Nager, Dr. Richard
Luxmoore and Professor
Bob Furness, for giving me this opportunity to undertake this
study. Without their
continued support and guidance this study would not have been
possible. Each of my
supervisors has contributed greatly and in individual ways to
provide critical, balanced and
informative encouragement. A particular thank you goes to Ruedi
for his endless patience
and continual review of my questions and queries. A sincere
thank you to each of you.
I also thank my funders: Biotechnology and Biological Sciences
Research Council,
and the National Trust for Scotland.
For continual support and help throughout the field seasons I
must thank my field
assistants: John McTague, Mark Wood, Kim Leyland and Hannah
Kippax. Thank you to
those people who agreed to look after cameras and nurse them
through the season: Will
Miles, Stacey Adlard, Newton Harper, Helen Moncrieff, and Mark
Newell. For permission
and help to install cameras onto puffin colonies, thank you to
Scottish Natural Heritage, the
National Trust for Scotland, Fair Isle Bird Observatory, the
RSPB, and the Centre for
Ecology and Hydrology. Thank you to Tycho Anker-Nilssen for
supplying data from
Norway, and also to Mike Harris for his advice. Thank you also
to Alex Robbins who was
willing to drop her own fieldwork to go and count puffin burrows
for me.
Thanks go to Josephine Pemberton for her kind donation of wool
samples from
Soay sheep and thanks to Gina Prior and Jill Pilkington for
their help whilst I stayed on
Hirta.
A particular thank you also goes to Jonathan Grant for his help
and company both
on and off Mingulay. A lot of knowledge was passed on about the
wildlife and history of
Mingulay to which I am indebted. Thank you to Jackie as well for
her support and regular
gifts of cake and flapjack. Donald and Francis, who drove the
boat to Mingulay, both
deserve a large thank you for their patience and support in
ferrying endless equipment,
food parcels, letters and people onto Mingulay. Without their
generosity life on Mingulay
would have been very different and difficult. Thank you as well
to Susan Bain and the
National Trust for Scotland in general for providing ample
opportunity and logistical
support for all of this work to happen.
-
9
To those people who were with me on Mingulay and provided
support, company
and friendship thank you very much: Gemma Jennings, Rob Dunn,
the Nager family and
Sophie Eastwood.
To everyone at the University of Glasgow who has provided
support throughout
this project, thank you very much. To everyone in Room 517,
thank you for advice, regular
tea breaks and help in general.
To all my friends in general who have supported me, read through
drafts and given
advice through these four years, thank you very much.
Finally, a very big thank you to all my family, in particular
Mum, Dad and Sam,
who have all provided encouragement, gifts of food on Mingulay
and continued support.
-
Chapter 1
Introduction
-
Chapter 1 Introduction
11
The study of island ecology has been the focus of scientific
research since the time of
Darwin and Wallace (Darwin 1859, Wallace 1881, Simberloff 1970).
Islands have
provided a natural system for insights into speciation,
evolution, competition and ecology
in general (MacArthur and Wilson 1967, Vitousek et al. 1996,
Mulder et al. 2011). These
developments have led to the practice of conservation and
furthered public concern
governing environmental matters and awareness (Whittaker and
Fernández-Palacios 2007).
A growing concern over the loss of biodiversity (Donald et al.
2001, Brooks et al.
2002) has been met with an increased understanding of the
importance of islands in
maintaining biodiversity, but there are many other aspects of
island ecology that need to be
understood (Mulder et al. 2011). Islands frequently contain
endemic species or populations
of species that have persisted and/or evolved in isolation from
mainland populations
(Bellingham et al. 2010, Jones 2010). Seabirds are one group of
species which are found
on islands around the world and are considered to be more
threatened than any other group
of birds with similar numbers of species (Mulder et al. 2011,
Croxall et al. 2012). The
decline in seabird populations and eradication of whole island
colonies has provided
ecologists an understanding of how island ecosystems are both
connected and dependent
upon the presence of seabirds.
The influence of seabirds on island ecology 1.1
Globally, there is an increasing understanding of the influence
seabirds have on the island
habitats they typically occupy. More often than not, this
influence only becomes evident
with their removal (Jones 1992, Croll et al. 2005, Maron et al.
2006, Grant-Hoffman et al.
2010) which often arises from the introduction of invasive
predators (Towns et al. 1990,
Bellingham et al. 2010). Removal of seabirds or a decline in
their population can result in a
shift of ecosystem structure and function (Maron et al. 2006).
The influence of seabirds on
land is however also evident without their extermination by the
noticeable signs of an
active seabird colony: the presence of burrows and nests,
altered vegetation, and often
most noticeably from guano (Ellis 2005).
Seabird colonies often influence the ecological communities
surrounding them and
these changes are well documented in the literature (Mulder et
al. 2011). Whilst most
-
Chapter 1 Introduction
12
studies document an immediate and direct influence to
surrounding vegetation (Norton et
al. 1997, Stapp et al. 1999, Polis et al. 2004, Ellis 2005,
Grant-Hoffman et al. 2010), there
are fewer studies which document indirect influence to other
organisms and trophic levels.
The following sections therefore detail firstly the direct
influence which seabirds have on
islands from three main processes: nutrient input, physical
disturbance and seed dispersal.
Secondly, a summary of the indirect influences which seabirds
have on island communities
is reported.
1.1.1 Nutrient input
Seabirds, as central-place foragers during the breeding season
(Fagan et al. 2007), will
forage at sea then return to their breeding island. These return
journeys result in the
transportation of nutrients from the marine environment to land,
which can have significant
effects upon the recipient community (Mulder et al. 2011).
In general, trans-boundary movement of predators, prey and
nutrients has been
shown to strongly affect local population dynamics and community
structure (Polis et al.
1997). Seabirds have been studied for their role in acting as a
vector of spatial subsidies
from marine to insular systems and the corresponding effect on
plant and animal
populations (Gillham 1956, Polis and Hurd 1996, Anderson and
Polis 1999, Sanchez-
Piñero and Polis 2000, Croll et al. 2005, Ellis 2005, Caut et
al. 2012). Seabirds act as
vectors through their capacity to introduce, often through
guano, large quantities of
marine-derived nutrients, such as forms of organic carbon,
phosphorous e.g. PO43-
, and
nitrogen e.g. NO3- and NH4
+ (Ellis 2005). Nutrients are further input through eggs
(Siegfried et al. 1978), feathers (Williams and Berruti 1978,
Smith 2008), carrion
(Williams et al. 1978, Ellis 2005), drip from the nasal salt
glands (Sobey and Kenworthy
1979), and from pellets and carcasses of chicks and adults
(Williams et al. 1978). This
represents a significant transfer of chemical elements of marine
origin into the terrestrial
environment.
Within studies on trans-boundary input nitrogen (N) and
phosphorous (P) are the
most commonly studied elements, due to the confounding
production of autochthonous
organic carbon from vegetation (Mizutani et al. 1991). N and P
are often deposited in such
-
Chapter 1 Introduction
13
large quantities that a comparison from studies on nutrient
input (Allaway and Ashford
1984, Anderson and Polis 1999) show deposition by seabirds far
outweighs agricultural
fertiliser application in the northern hemisphere (Pearson and
Stewart 1993, Rajakaruna et
al. 2009).
It is important to note that although seabirds mediate change in
soils (Heine and
Speir 1989, Jakubas et al. 2008, Schaeffer et al. 2008, Wright
et al. 2010), they also affect
other systems within insular ecosystems indirectly. For example,
excess ammonia and
nitrates within soils under the right conditions can lead to
ammonia volatilisation (Pearson
and Stewart 1993). Volatilised ammonia can then enter the
atmosphere in vast
concentrations and be subsequently deposited back onto land
through wet and dry N
deposition (Wright et al. 2010), leading to further
acidification, eutrophication and impacts
on nearby soils and plant communities (Fangmeier et al. 1994,
Wilson et al. 2004b).
The transfer of nutrients by seabirds onto land generally
increases primary
productivity, although this is dependent upon the nutrient’s
concentration and chemical
composition (Polis et al. 1997, Smith 2008, Rajakaruna et al.
2009). A comparison
between areas affected by birds and those unaffected can show
floral composition to differ
slightly or even be drastically changed as a result of guano
deposition (Wait et al. 2005).
Furthermore, the provision of nutrients from seabird carcasses
or prey remains can support
scavenging organisms (López-Victoria et al. 2009).
1.1.2 Physical disturbance
Physical disturbance of soil by animals can influence ecosystem
structure and function
(Butler 1995, Whitford and Kay 1999, Bancroft et al. 2005a) but
studies showing such
effects from seabirds on island systems remain scarce (Butler
1995). Bancroft et al.
(2005b) suggest that soil processes on island systems are of
particular ecological concern
given that islands typically have soils of poor aggregation and
nutrient composition, high
water infiltration rate, harsh environmental conditions, and
communities, which exist in an
unstable state of equilibrium.
In general physical disturbance arises during the breeding
season, through the
processes of nest building and the actions of scratching,
pecking and also trampling of
-
Chapter 1 Introduction
14
plants and soils (Sobey and Kenworthy 1979, Mulder et al. 2011).
The exact influence of
nest building varies with each seabird species, dependent on
nest type, e.g. tree, burrow,
crevice or surface nests. The variables often found to be
influenced by physical disturbance
commonly include physical soil characteristics e.g. temperature,
bulk density, water
content, water repellency, infiltration and soil strength
(Bancroft et al. 2005b). Chemical
characteristics of soil influenced by physical disturbance
include altered nutrient
concentrations, soil pH and conductivity (Garcia et al. 2002,
Bancroft et al. 2005b).
Overall though, there are relatively few studies which have
demonstrated the negative
impacts of disturbance by non-herbivorous vertebrates (Maesako
1999).
Between seabird species, it is considered that burrow nesting
species could have
one of the largest effects on land through physical disturbance
(Mulder et al. 2011),
although the degree of influence depends upon nesting density,
the age of the colony and
the burrowing species (Mulder et al. 2011). In general,
burrowing species typically of the
order Procellariidae, e.g. shearwaters and petrels, and
Charadriiformes of the family
Alcidae, e.g. puffins, are known to burrow extensively.
Burrowing reduces the stability of
the top soil (Boag and Alexander 1996) and alters physical and
chemical properties of
soils, which can alter species richness and seedling growth
(Roberts et al. 2007, Grant-
Hoffman et al. 2010). The impacts of burrowing and trampling
further affect communities
through the exposure of roots and from initiating retrogressive
succession (Maesako 1991,
1999).
By comparison surface nesters can have a relatively small impact
on islands,
although this is again dependent upon each species and their
nesting density (Mulder et al.
2011, Kolb et al. 2012). Sobey and Kenworthy (1979) showed that
treading and sitting by
herring gulls Larus argentatus caused mechanical damage to
plants and suppressed
vegetation in regularly frequented areas. Boundary clashes
between birds may also result
in indiscriminate disturbance through pecking and pulling of
both dead and alive plant
material (Mulder et al. 2011). The impact of tree nesters varies
with each species. Larger
species such as colonial cormorants or frigatebirds can cause
extensive damage to
vegetation during nest building, landing and taking off (Mulder
et al. 2011). Other species
such as the marbled murrelet Brachyramphus marmoratus are
relatively smaller and
consequently cause very little physical disturbance (Piatt et
al. 2006, Mulder et al. 2011).
Nest building and use, in general, can result in the suppression
of plant growth within the
-
Chapter 1 Introduction
15
area due to the activities of forming the nest, incubation and
chick development (Mulder et
al. 2011).
1.1.3 Seed dispersal
There are several documented cases of seabirds initiating
vegetative changes on islands
due to the dispersal of seeds (Gillham 1956, Wait et al. 2005,
Ellis 2005, Sekercioglu
2006, Mulder et al. 2011). Seabirds have become such a prolific
and effective vector of
seeds that many species of plant have developed obvious features
to facilitate this
mechanism (Howe and Smallwood 1982, Burger 2005), e.g. cypselae
of Asteraceae
commonly have viscid extensions which enhance adhesion to
animals (Sorensen 1986).
These adaptations to birds have resulted in greater dependence
on certain taxa for dispersal
than others, e.g. a study in the Great Lakes showed that 61% of
the flora is adapted to
dispersal by birds.
This adaptation favouring dispersal by birds often facilitates
the spread of alien
species (Gillham 1956). In a review by Ellis (2005) it was shown
that the majority of
studies documented a greater proportion of cosmopolitan species
compared to native plant
species within seabird colonies (Morton and Hogg 1989, Vidal et
al. 2003). For example,
seabird colonies can often contain cereals and arable weeds
(Gillham 1956). The
introduction of alien and cosmopolitan species by seabirds onto
islands can result in the
disappearance of endemic species and consequent ecological
perturbations within the
original phytocenosis (Vidal et al. 1998, Ellis 2005).
Furthermore, the diversity of alien
plants and plants capable of producing bird-dispersed seeds is
closely correlated with
island area size (McMaster 2005) and colony density (Heatwole
and Walker 1989),
respectively. Overall, it is well established that turnover
rates on islands with seabird
colonies are notably higher than those without, due to propagule
dispersal and patch
disturbance (Vidal et al. 2000).
-
Chapter 1 Introduction
16
1.1.4 Indirect influence
Seabirds further influence terrestrial systems by indirectly
interacting with island
communities through predominantly promoting bottom-up dynamics
(Sanchez-Piñero and
Polis 2000). The provision of nutrients, compared to physical
disturbance or seed dispersal,
can be viewed as potentially the largest factor driving
community dynamics on seabird
islands. For example, the presence of seabirds and their guano
can alter whole island
systems by fuelling communities of lizards and crabs
(López-Victoria et al. 2009),
increasing beetle abundance (Sanchez-Piñero and Polis 2000),
altering soil microbial
systems (Wright et al. 2010), increasing algal production
(Bosman and Hockey 1986) and
also altering behaviour of herbivorous mammals (Jakubas et al.
2008).
In general, there is an increasing number of studies which
document cross-scale
interactions of organisms between ecosystems, which can alter
and impact upon ecosystem
processes and community dynamics (Polis et al. 1997, Knight et
al. 2005, Peters et al.
2007, Caut et al. 2012, Bultman et al. 2014). Studies have shown
how a range of species
influence ecosystems across apparent ecological boundaries, for
example: emergent
aquatic insects can subsidise terrestrial systems (Bultman et
al. 2014), migrating pacific
salmon can fuel inland terrestrial populations (Cederholm et al.
1999, Adams et al. 2010)
and sewage discharge can increase aquatic insect populations
leading to provision of
nutrients for swallows (Wayland and Hobson 2001). These studies
highlight the
importance of nutrient transfer across ecological boundaries and
give insight into
ecosystem functioning.
In the context of seabirds, understanding the extent to which
terrestrial systems are
impacted by seabirds is of critical importance to the
conservation of insular biodiversity
(Caut et al. 2012). However, the influence which seabirds have
on terrestrial systems does
not consider nutrient transfer in detail from seabirds to
mammalian herbivores or
passerines. Understanding the processes underpinning nutrient
transfer throughout islands
is of significant conservation value given the high level of
species endemism found on
islands globally (Ellis 2005, Bellingham et al. 2010, Jones
2010)
Overall, seabirds can be seen to increase habitat heterogeneity
by altering plant
species richness (Vidal et al. 1998) and the probability of
plant invasion (Mulder et al.
2009). However, at high nest and burrow density plants may not
be able to survive the
-
Chapter 1 Introduction
17
severe disturbance (Vidal et al. 2000). The transfer of vast
quantities of nutrient onto land
from seabirds can be seen to drive primary productivity, which
can impact upon higher
trophic levels (Mulder et al. 2011). It is important to note
that seabirds breed within a
limited period in the year, so their effects on land are
temporally variable (Anderson et al.
2008). Importantly, however, is that the effects of seabirds on
land are mediated by
population size. With an increasing population size and density
their effects on land
increase concomitantly (Mulder et al. 2011, Kolb et al. 2012)
Therefore, to fully
understand how seabirds influence islands, reliable data on
seabird population sizes,
collected via appropriate monitoring schemes is required.
Monitoring 1.2
There are several reasons and techniques for counting birds
(Jones 1992, Montevecchi
1993, Walsh et al. 1995, Anker-Nilssen et al. 1996, Bertram et
al. 1999, Bibby 2000).
Choosing an appropriate method is key to provide accurate
records of population size,
which can provide in part an insight into the influences of both
natural and anthropogenic
change on seabirds (Bellingham et al. 2010). Counting colonially
breeding seabirds
however poses certain issues, which must be determined prior to
any study. These issues
include, but are not constrained by, the problems of estimating
the proportion of breeding
to non-breeding birds, locating and counting birds in remote,
difficult terrain, evaluating
the proportion of the population feeding away from the colony
and the influence of
weather on colony numbers (Bibby 2000).
The choice of monitoring scheme is dependent upon each species
with a variety of
methods suggested for both within and across species. For
relatively cryptic species, such
as burrow or crevice nesters, then determination of nesting
location can be subjective,
unless direct confirmation of the presence of eggs or chicks is
possible (Anker-Nilssen and
Røstad 1993, Walsh et al. 1995). Other methods for monitoring
cryptic species range from
knock-down tags to determine burrow occupancy to radio telemetry
for locating nests
(Gaston et al. 1988, Kenward 2001). Counts of more visible
species either use numbers of
nesting locations or apparently occupied sites, given that
determination of actual
occupancy can be difficult with distance or densely nesting
species (Walsh et al. 1995).
-
Chapter 1 Introduction
18
Monitoring of seabirds must also consider the temporal
variability in abundance at
different scales. Many studies have shown how seabird abundance
varies over the season
and this variation must be understood prior to any monitoring
scheme (Nettleship and
Birkhead 1985, Klomp and Furness 1990, Finney et al. 2003,
Regular et al. 2010, Harris
and Wanless 2011). For example, the erratic attendance of many
alcid species can limit the
reliability of counts of individuals as an estimate of
population size (Walsh et al. 1995,
Calvert and Robertson 2002a, Harding et al. 2005).
Each method of monitoring used often produces variable results
and the
discrepancy between monitoring methods elucidates the need for a
meaningful,
comparable and standardised method (Nettleship and Birkhead
1985). Some methods of
monitoring are also known to adversely affect seabird
populations by the continued
presence of observers, which can promote nest abandonment, in
particular during the early
incubation stage (Gotmark 1992, Rodway et al. 1996,
Albores-Barajas et al. 2009, Shoji
and Gaston 2010). Monitoring seabirds without human disturbance
is of importance in
maintaining and recording natural breeding success.
Monitoring schemes often have logistical constraints, which can
limit the scope of
recording (Dickinson et al. 2008). Counting seabirds manually,
often in remote locations
under harsh conditions, can become impractical and/or incur
significant costs of labour and
time (Anker-Nilssen et al. 1996, Trathan 2004, Dickinson et al.
2008). Consequently,
automated approaches to monitor seabirds have been developed to
negate some of the
limitations from manual monitoring. In particular, the use of
photography has become
prevalent within wildlife monitoring in general (Martin and
Geupel 1993, Cutler and
Swann 1999, Gula et al. 2010) and also for seabirds (Wanless and
Harris 1986, Hatch
2002, Huffeldt and Merkel 2013). The use of photography to
monitor seabirds has been
undertaken from ground-based (Mudge et al. 1987, Harding et al.
2005), aerial (Buckland
et al. 2012) and even satellite platforms (Lynch et al. 2012).
However, there are greater
costs involved with photographic approaches not based on the
ground (Lynch et al. 2012).
Also, for aerial platforms there is the potential for
disturbance to seabirds, for example
from low flying planes (Trathan 2004, Buckland et al. 2012).
Limitations for ground-based
photography include a need to access and find vantage points to
photograph a colony, but
for some scenarios, ground-based photography is the only
sensible option if photography is
to be used. Ground-based photography can be used for monitoring
species which nest on
-
Chapter 1 Introduction
19
cliffs and/or in burrows and whose detectability is reduced when
monitored with aerial or
satellite platforms.
Time-lapse photography provides one way in which to monitor
seabird populations
by repeated photographs across a period of time with relatively
low cost, providing
detailed insight into seabird populations (Lorentzen et al.
2012, Huffeldt and Merkel
2013). Although technology has become possible for long-term
monitoring using
photography, the approach has not been fully explored for its
potential (Huffeldt and
Merkel 2013). There is a need to explore long-term monitoring of
seabirds using time-
lapse photography to examine changes in population size.
However, whenever monitoring
is undertaken there is a need to ensure that the same approaches
are undertaken over time
for comparability and that any change in methodology considers
sampling limitations and
error between different methods (Buckland et al. 2012)
Overall, knowing the size of seabird populations is of
importance, not least for
understanding their effect on island systems, but also for their
close association with the
marine environment. This association has allowed seabirds to be
viewed as indicators of
the state of the oceans’ health (Harris and Bailey 1992,
Montevecchi 1993). The ability of
seabirds to exert top-down control within trophic networks and
also respond to bottom-up
influences makes them important ecological bioindicators (Fort
et al. 2011). Furthermore,
birds can be viewed as bioindicators as the concept is readily
understood by the public and
policy-makers, due to their popularity and often iconic status
(Crick 2004). Using seabirds
as bioindicators allows a wide-spread understanding of the
survival, recruitment and
productivity of seabirds, which is imperative for any initiation
of conservation and
management (Nettleship and Birkhead 1985). Seabirds can also
allow for an indication of
prey abundance, whereby a particular choice of prey can be a
short-term indicator of prey
availability and allow inferences to be made on the health of
the marine ecosystem
(Montevecchi 1993). Equally, the numbers of non-breeding
seabirds could provide a
sensitive indication to the effects of environmental change,
since non-breeders act as a
buffer to such stress (Klomp and Furness 1992).
-
Chapter 1 Introduction
20
Project rationale and study species 1.3
Within the United Kingdom, there are over seven million breeding
adults of 25 species of
seabird (Lloyd et al. 2010, Wanless et al. 2010) and within
Scotland, and the islands of the
Outer Hebrides and Shetland in particular, seabirds are a common
part of the fauna.
Consequently, seabirds on these islands can have a significant
role in ecosystem processes
and community structure. However, there are relatively few
studies examining the
influence of seabirds on islands in Scotland with most studies
focussing upon systems in
the Pacific (Anderson et al. 1998, Anderson and Polis 1999,
Maron et al. 2006) and the
islands around New Zealand (Hawke and Newman. 2004, Roberts et
al. 2007, Bellingham
et al. 2010). Consequently, there is a need to explore the
influence of seabirds on islands in
more temperature latitudes, such as in Scotland.
This study was limited to two study species in Scotland: the
great skua Stercorarius
skua and the Atlantic puffin Fratercula arctica. These species
were chosen for their
different nesting habits, their variable nesting density and
different behaviours, which may
all affect their influence on island systems. Given the
importance of understanding
population size a novel method for monitoring Atlantic puffins
is trialled using time-lapse
photography. Studies were conducted on three islands around
Scotland: Mingulay and
Hirta (St. Kilda), Outer Hebrides, and Fair Isle, Shetland.
To discuss each species in further detail, the great skua is a
large generalist
predator, which inhabits islands across the North Atlantic from
Iceland east to Svalbard
(Lloyd et al. 2010). The great skua is a surface nester
typically found on moorland close to
the sea with an estimated global population of c. 16,000 pairs,
of which 9,600 pairs are
within Scotland (Mitchell et al. 2004, Lloyd et al. 2010). Their
influence on land is
however not well documented with only limited observations on
their immediate influence
to the colony surface (Engelskjon 1986, Furness 1987, Klomp and
Furness 1990). The
ability to monitor skuas is relatively easy with estimates of
the number of pairs derived
from visible signs of territory occupancy on the colony surface
or also in part by their
relatively aggressive nature (Furness 1987, Walsh et al.
1995).
Secondly, the Atlantic puffin is a small species of auk with a
distribution from the
East coast of North America to the Northern coast of Russia with
a global population
estimated up to 6 million pairs (Lloyd et al. 2010). The
Atlantic puffin is a colonial
-
Chapter 1 Introduction
21
burrow-nesting seabird, with colonies scattered around the UK,
although the largest
colonies and most of the UK population is in Scotland (Lloyd et
al. 2010). Atlantic puffins
create their burrows in suitable substrate or they may nest in
crevices and boulder scree.
This subterranean habit combined with their erratic colony
attendance makes monitoring
Atlantic puffins complicated (Nettleship 1976, Walsh et al.
1995). The influence which
Atlantic puffins have on land has only been documented for their
immediate influence on
the colony but their wider impact on island systems is poorly
understood (Boag and
Alexander 1996, Harris and Wanless 2011).
Project aims 1.4
The following chapters of this thesis are thus structured to
primarily explore the effect of
Atlantic puffins and great skuas on islands in Scotland. An
understanding of the population
size is critical to predicting the effect of Atlantic puffins on
land. Therefore, an alternative
method to monitor Atlantic puffins using time-lapse photography
is trialled.
Chapter 2 aims to determine how the attendance of the Atlantic
puffins varies over
different temporal and spatial scales and in response to
environmental variables. However,
rather than using direct observations of puffin attendance this
method employs time-lapse
photography. Cameras are installed across a wide geographical
range to explore variation
in attendance within and between colonies, with consequences for
methods in monitoring.
Chapter 3 aims to show how time-lapse photography can be used to
monitor
Atlantic puffins directly as a stand-alone approach. This study
gives further consideration
to sampling intensity and the relationship between the numbers
of puffins observed and the
actual breeding population. Finally, a novel approach of
associating the spatial position of
Atlantic puffins in relation to burrows is studied.
Chapter 4 estimates the mass of nutrients that a population of
Atlantic puffins and
great skuas brings onto the island of Mingulay, Outer Hebrides,
Scotland, during the
breeding season. Estimates of nutrient input focus upon nitrogen
and phosphorous and are
calculated using a bioenergetics approach.
-
Chapter 1 Introduction
22
Chapter 5 explores how Atlantic puffins and great skuas
influence island ecology
in Scotland. The study looks at plant community diversity and
also variation in nitrogen
within grasses over different islands in response to seabird
presence. The study also aims
to describe how nutrients are transferred from seabirds to
mammalian herbivores and how
this may influence their populations.
Chapter 6 investigates how nutrients are moved around different
components of
island systems. This study specifically looks at how nutrients
are moved from a large
colony of common guillemots Uria aalge and razorbills Alca torda
to a large passerine
population of house martins Delichon urbica in the Baltic
Sea.
Chapter 7 brings the thesis together in a General Discussion,
which ties how seabirds
influence island ecology with a potentially wide reaching issue
of conservation. The
applicability of using time-lapse photography is further
discussed.
-
2 Chapter 2.
Variation in Atlantic puffin Fratercula arctica
colony attendance; monitoring using time-lapse
photography
-
Chapter 2 Attendance patterns of the Atlantic puffin
24
Abstract
Attendance, the observable numbers of seabirds at a colony, is
often used to reflect
population size. Obtaining detailed of counts of seabirds to
determine attendance requires
access to monitor colonies, which are often in places that are
logistically difficult to access
and have harsh environmental conditions. Consequently,
significant costs time and labour
can be incurred in determining attendance. The Atlantic puffin
Fratercula arctica is a
species where detailed counts of attendance are required given
it has both a notoriously
erratic pattern of attendance and also a subterranean presence.
An understanding of those
factors influencing attendance needs to also be considered if
attendance is used for
purposes of monitoring. Time-lapse photography is proposed as a
novel method with
which to obtain high resolution temporal data on puffin
attendance across the season, with
minimal costs of installation and processing. For this study,
time-lapse cameras were
installed on islands around Scotland and demonstrated that
detailed attendance patterns of
the Atlantic puffin can be detected using time-lapse
photography. The cameras showed that
puffins exhibited a general pattern of increasing attendance
throughout the day, with a
cyclic pattern of attendance across days with a periodicity of
around 5 to 7 days. Puffin
attendance was shown to significantly decrease with spring tides
and one colony showed
decreased attendance with increasing wind speed. There was no
observable effect of
temperature on puffin attendance and predator presence caused
decreased attendance. The
use of high temporal resolution data to define periods of
attendance lends support to the
possibility that counts of puffins could be used for purposes of
monitoring.
-
Chapter 2 Attendance patterns of the Atlantic puffin
25
Introduction 2.1
Monitoring seabirds is of importance to establish how the size
of each species’ population
changes over time. The inextricable link between seabirds and
the marine environment has
resulted in seabirds being viewed as indicators of the oceans’
health (Montevecchi 1993,
Lloyd et al. 2010). Estimating the size of a seabird population
can be undertaken using the
number of individuals observable in a colony, termed attendance
(Walsh et al. 1995,
Calvert and Robertson 2002a). Attendance can however vary
markedly over time, both
within and between days, which has implications when estimating
population size from
counts of individuals (Slater 1976, Cairns 1979, Hatch 2002,
Harding et al. 2005, Huffeldt
and Merkel 2013). Understanding the variability in colony
attendance is therefore of
importance if indices of population size are to be reliably
derived.
The Atlantic puffin Fratercula arctica (hereafter termed
“puffin”) is an example of
a colonial, burrow-nesting seabird that breeds in both burrows
and crevices, and which is
known to be notoriously variable in its colony attendance
(Brooke 1972, Calvert and
Robertson 2002b, Harris and Wanless 2011). Studies have shown
how attendance varies on
both a daily and seasonal scale. On average daily attendance
increases after midday and
peaks from the late afternoon to evening, although day length
can also influence attendance
(Corkhill 1971, Harris and Wanless 2011). Over a seasonal scale
puffins can have cycles of
attendance with variable periodicity. Periodicity can vary from
7 days as on Great Island,
Canada (Nettleship 1972), to 4 to 11 days as on Lovunden, Norway
(Myrberget 1979,
Harris and Wanless 2011), although cyclicity is often around 5
days (Corkhill 1971, Lloyd
1972, Calvert and Robertson 2002a, Harris and Wanless 2011).
Different colonies have
also been found to be both synchronous and asynchronous in their
attendance within and
between islands (Corkhill 1971, Harris and Wanless 2011). In
addition, attendance can
vary across the season with breeding stage and also the arrival
of non-breeding prospecting
puffins (Lloyd 1972, Cairns 1979, Harris and Wanless 2011).
Attendance of puffins, and seabirds in general, can also be
influenced by
environmental factors (Walsh et al. 1995, Calvert and Robertson
2002a, Harris and
Wanless 2011). Attendance for many species of, for example
alcid, has been shown to be
differentially influenced by atmospheric conditions. Studies
have shown how increased
wind speed decreases attendance by reducing flight efficiency
(Lloyd 1972, Hatch 2002,
Calvert and Robertson 2002a), although some studies find no
significant effect (Blet-
-
Chapter 2 Attendance patterns of the Atlantic puffin
26
Charaudeau et al. 2010). Equally, some studies find no influence
of temperature on
Atlantic puffin attendance (Calvert and Robertson 2002a), whilst
others do (Blet-
Charaudeau et al. 2010). Furthermore, wind direction,
precipitation and cloud cover are
observed to have little influence on attendance (Hatch 2002,
Calvert and Robertson 2002a).
Tidal conditions may also influence colony attendance by its
effect on prey
availability and consequently foraging behaviour (Piatt et al.
1990, Irons 1998). As the
magnitude of tidal oscillations increases with both daily tides
and longer spring/neap tidal
cycles, seabird prey, e.g. zooplankton and their prey, may
become concentrated at tide rips
and offshore fronts (Piatt et al. 1990). Studies have shown an
effect of tidal state on colony
attendance for a range of seabird species (Slater 1976, Piatt et
al. 1990, Vermeer et al.
1993, Irons 1998). However, other studies show no effect of tide
on attendance or even
contradict prior studies (Piatt and McLagan 1986, Piatt et al.
1990), though this may be a
result of variation in local tidal range (Cairns 1979). There
are limited studies looking at
the variation in puffin attendance with regards to tides
(Corkhill 1973), although there have
been contradicting studies on the interaction of puffins with
tidal state (Wanless et al.
1990, Robbins 2012).
Although not considered an external factor influencing
attendance, the extent of
colony attendance may be in part governed by how long individual
puffins spend on land.
Previous studies have shown that the length of time a puffin
spends on land is correlated
with colony attendance (Calvert and Robertson 2002a). There may
be an advantage to
spending more time on land when surrounded by conspecifics to
reduce the chance of
predation (Calvert and Robertson 2002a).
In general, the range of contradicting results for the influence
of environmental
variables on puffin attendance requires further study. The
variation in attendance and
influence of environmental variables may be colony specific
(Harris and Wanless 2011).
Understanding which factors influence puffin attendance can lead
to further improvements
in the reliability of counts of individuals, as an index of
population size.
For puffins, estimates of colony attendance are often undertaken
by manually
counting in the field the numbers of individuals present
(Corkhill 1971, Lloyd 1972,
Calvert and Robertson 2002a). This method can however take
significant time to undertake
and consequently attendance is often monitored at a lower
resolution and/or frequency
-
Chapter 2 Attendance patterns of the Atlantic puffin
27
(Myrberget 1979). This limited monitoring of puffins can
therefore lead to an
unrepresentative estimate of the population size given the high
variation in attendance.
Equally, very few studies look at attendance in detail across
colonies and time, which is
likely to be due to the significant man-hours and cost involved
in manually recording
attendance at multiple colonies (Corkhill 1971, Lloyd 1972). An
automated approach to
monitor puffins could allow high resolution temporal data to be
collected across the
breeding season and between colonies, providing detailed
information on the variability in
attendance.
Time-lapse photography provides a way in which to record puffin
attendance
automatically over time. Time-lapse photography and video are
both methods which have
been used in avian applications, such as in studies of behaviour
and breeding success
(Dickinson et al. 2008, Lorentzen et al. 2010, 2012, Gula et al.
2010, Huffeldt and Merkel
2013). Studies have also used time-lapse to count individuals of
different seabird species
for purposes of monitoring (Harris and Wanless 1983, Mudge et
al. 1987, Piatt et al. 1990,
Zador and Piatt 1999, Harding et al. 2005). The use of
time-lapse photography to monitor
puffins is however limited to one study, which demonstrates how
productivity of puffins
can be inferred from attendance (Anker-Nilssen 2010).
In general, time-lapse photography could be used in remote
locations with visits
only to install and collect the camera system at the beginning
and end of the season,
respectively (Huffeldt and Merkel 2013). The increased
commercial viability of camera
systems means that the use of time-lapse photography could
become competitive with
visual methods for monitoring (Buckland et al. 2012, Huffeldt
and Merkel 2013).
Automated photography can furthermore be used to collect data on
attendance at a high
temporal resolution with minimal cost and effort from remote
areas.
The aims of this study were to examine the attendance of
Atlantic puffins over
daily and seasonal scales, using time-lapse photography at a
resolution not recorded before.
The study also aimed to understand the effect of environmental
variables on puffin
attendance. Understanding puffin attendance has implications for
monitoring populations.
The installation of cameras on islands around Scotland aims to
also give an indication of
the geographical variation in attendance within and between
colonies.
-
Chapter 2 Attendance patterns of the Atlantic puffin
28
Methods 2.2
This study was conducted on five Scottish puffin colonies; three
colonies on Mingulay
(termed “Original”, “Otter” and “Arnamul”), Outer Hebrides, one
colony on Fair Isle and
one colony on Unst, Shetland (Hermaness) (Figure 2.1). Data were
collected over 2011 and
2012 for the two colonies: Original on Mingulay, and Fair Isle.
In 2012 two more sites
were added on Mingulay: Otter was within the same larger,
cliff-top colony on the east
side of Mingulay as Original but the two observation areas were
separated by c. 150 m and
the census areas did not overlap. Arnamul was a colony on the
west of Mingulay on the
side of a stack, c. 100 m offshore and c. 2.5 km from the
Original and Otter colonies.
Hermaness is c. 600 km North-East from Mingulay and was located
on the side of a cliff
whereas Fair Isle is c. 150 km South of Hermaness and
represented a cliff-top colony.
Multiple colonies and years were chosen in order to examine
variation in inter-colony and
inter-year attendance patterns.
Because hatching dates in different puffin colonies around
Scotland within the
same year can be similar (Harris 1982, Harris and Rothery 1985,
Mavor et al. 2006) the
median hatching date observed on Mingulay (11th
June in both years, pers. obs.) was used
to split the study period into a pre-hatching (incubation) and
post-hatching (chick-rearing)
stage. Observations on Fair Isle started on 1st July in 2011 and
ended on 1
st June in 2012
and thus would be unlikely to cover the pre- and post-hatch
stage, respectively.
Observations on Hermaness started on the 14th
June 2012 and these data were considered
to come from the post-hatch stage only, although admittedly this
may be within the range
of hatch dates. Thus, five of the seven datasets contained data
for the pre-hatch stage and
of these five datasets four also had data for the post-hatch
stage (Table 2.1). The post-
hatch stage has a further two datasets from colonies not studied
in the pre-hatch stage, thus
a total of 6 datasets.
2.2.1 Counts of puffins using photography
Colony attendance at each site was established using time-lapse
photography. At each site
a camera with an attached timer device was positioned and set to
take photographs over
time. A digital single-lens reflex camera (Canon 500D or 550D)
with a zoom lens (Canon
EF-S 18-55 mm) was used with a minimum 8 GB SD card for storing
photographs. A
-
Chapter 2 Attendance patterns of the Atlantic puffin
29
simple unbranded timer device was fitted to the camera, which
allowed the camera to take
repeated photographs (time-lapse photography) with a
user-defined interval. The interval is
defined as the time elapsed between successive photographs. The
camera and timer device
were then installed in a custom-built weatherproof housing. The
cameras were located to
maximise the field of view of the puffin colony, whilst
considering image resolution. The
final positioning was a balance between distance from colony,
topography and underlying
substrate.
The interval set on the timer device varied between colonies and
year, to explore
the effectiveness of different sampling intervals and also to
fit with logistical constraints of
the time available for camera maintenance by volunteers. On
average the battery and SD
card were changed every two to seven days, depending upon
photographic interval. A
minimum of 10 minutes was intended as the sampling interval;
however due to storage
limitations or if maintenance visits could not be carried out
with high frequency then the
sampling interval may have been longer (Table 2.1).
The periods the cameras operated covered mid-late incubation to
late chick-rearing.
On Mingulay, cameras were installed and set-up from around
mid-May until mid-July,
coinciding with logistical availability. The study duration on
other colonies was defined by
volunteer availability. Cameras were installed and left to run
between 17 and 60 days
(Table 2.1). Photographs were taken when light levels allowed
for correct exposure; thus
images at night are often excluded. Due to camera malfunctions
there were periods of time
when the cameras were not working, which reduced the effective
temporal coverage
(Table 2.1). Furthermore, a small percentage of individual
photographs could not be used
(mean = 1.4 % ± 1.9 SD) when, for example, adverse weather
obscured the lens (Table
2.1). For analyses, these data are assumed missing at random and
those periods separated
by more than one missing day are treated as separate time
series. Hermaness was the only
colony which had its time series split based upon periods of
missing data. Two periods of
camera malfunction on Hermaness, which lasted 10 and 8 days
respectively, resulted
consequently in three shorter time series. Two of the shorter
time series were less than 9
days each and were discarded, whilst the longer time series
retained and used within this
study was 17 days long.
At the end of the season, the numbers of puffin present on the
ground were visually
counted on each usable photograph. A sample of photos from
Original in 2012 was
-
Chapter 2 Attendance patterns of the Atlantic puffin
30
independently counted by two different observers and showed a
high correlation between
observers (Pearson’s correlation = 0.98, n = 133, P < 0.001).
The numbers of observable
puffins on the ground represented a measure of colony
attendance.
2.2.2 Attendance patterns
Counts of attendance over different times of day can be very
variable (Brooke 1972, Harris
and Wanless 2011). To observe the average change in daily
attendance the numbers of
puffins over time were aggregated based upon two daily time
scales, following Calvert and
Robertson (2002a).
Firstly, to illustrate the daily attendance pattern at
relatively fine temporal
resolution the mean number of puffins within every 30 minute
interval across the day was
calculated. The mean number of puffins was taken from averaging
the number of puffins
each day across the season into 30 minute intervals from
midnight. However, an analysis
exploring variation in attendance between times of day using 30
minute intervals would
provide too many levels within a factor for interpretable
analysis. Thus, the day was split
into five equal periods of ~ 5 hours: Early Morning (00:00 to
< 04:48 h), Morning (04:48
to < 09:36 h), Afternoon (09:36 to
-
Chapter 2 Attendance patterns of the Atlantic puffin
31
2.2.3 Factors related to attendance
2.2.3.1 Environmental variables
Environmental variables were related to colony attendance to
understand what factors
potentially influence the numbers of puffin seen on land. Data
on daily weather for
Mingulay were obtained from Tiree (63 km away). Although there
is a closer weather
station (South Uist, 57 km away), Tiree’s location is
geographically more similar to
Mingulay, in that the effect of topography on weather is
relatively less, than South Uist
with a relatively greater land mass and topographic relief
upwind (Mayes and Wheeler
1997, Owen et al. 2004). Weather data for Hermaness came from
the closest weather
station at Sella Ness, Shetland (48 km away). Meteorological
data (mean temperature (oC)
and mean wind speed (km h-1
)) for these weather stations were collected from Weather
Undergroud (2014). A strong, positive correlation between data
from Weather
Underground and limited data obtained from the Met Office for
Tiree (2nd
June to 15th
July
2011), ensured that the data from Weather Underground and the
Met Office were
comparable for that station (mean Pearson’s correlation (± 95 %
confidence interval) =
0.81 (0.70 – 0.89). To calculate the mean Pearson’s correlation
both weather time series
were first-differenced by subtracting the value at ti from ti+1
to remove any long-term trend
(Bjornstad et al. 1999). However, given serial correlation
within each time series data are
not independent (Buonaccorsi et al. 2001). A bootstrap 95 %
confidence interval for the
mean synchrony was produced from sampling (n = 1000) with
replacement between each
time series and recalculation of Pearson’s correlation
(Bjornstad et al. 1999, Buonaccorsi
et al. 2001). Bootstrapping provides a measure of significance
by randomisation, whilst
accounting for serial correlation (Bjornstad et al. 1999).
The behaviour and attendance of seabirds has been found to be
associated with the
magnitude of tidal oscillations, which could influence the
concentration of seabirds’ prey
(Slater 1976, Cairns 1979, Vermeer et al. 1993, Irons 1998).
Consequently, mean daily
tidal height was related to mean daily attendance of puffins to
observe any influence of
seasonal tidal rhythms, e.g. spring and neap tides, on
attendance. The resolution of tidal
data on a daily scale does not allow for examination of the
effect of within-day high and
low tides. The mean daily tidal height was obtained from Barra
Head (~ 3 km from
Mingulay) in 2011 and 2012 and Burra Firth (adjacent to
Hermaness) in 2012. All tidal
data were obtained from the UK Hydrological Office (2013). The
association between
-
Chapter 2 Attendance patterns of the Atlantic puffin
32
attendance, and meteorological and tidal data were not analysed
for Fair Isle given the
short study duration and long intermittent periods of missing
data (Table 2.1).
2.2.3.2 Predators
Original 2012 also had the presence and absence of the predators
of puffins recorded for
each photograph. Predators were defined as the great
black-backed gull Larus marinus, the
herring gull Larus argentatus, the lesser black-backed gull
Larus fuscus, the great skua
Stercorarius skua and the hooded crow Corvus cornix (Harris and
Wanless 2011). The
presence of predators was recorded to explore their effect on
the numbers of puffin present.
2.2.4 Statistical analyses
2.2.4.1 Daily attendance patterns
The change in daily attendance patterns between the five daily
periods was analysed using
a Generalised Linear Mixed Model (GLMM) with a Gaussian error
structure. The mean
number of puffins was firstly calculated for each of the five
daily periods each day. The
GLMM was used to analyse the difference in the mean number of
puffins between daily
periods and colonies. The fixed effects of period of day and
colony were used with an
interaction between the two terms. The GLMM was used with a
random effect of day of
year nested within colony to account for repeated measurements
over time and within
colonies. The model included data from the colonies of Mingulay,
Fair Isle and Hermaness
over both 2011 and 2012. Model structure: Mean number of puffins
~ daily period *
colony, random effect = day of year/colony.
A GLMM with a Gaussian error structure was also used to analyse
the difference in
the mean number of puffins between daily periods over two
different stages of the puffin’s
breeding cycle: pre-hatch and post-hatch. Fixed effects included
period of day, breeding
stage and colony with interactions between breeding stage and
colony, breeding stage and
period of day and a three-way interaction between all three
fixed effects. A random effect
of day of year nested within colony was included to account for
repeated measurements
-
Chapter 2 Attendance patterns of the Atlantic puffin
33
over time and within colonies. Only colonies from Mingulay were
analysed given
applicable data over both the pre- and post-hatching stages.
Model structure: Mean number
of puffins ~ breeding stage * daily period breeding stage *
colony + breeding stage *
colony * daily period, random effect = day of year/colony.
2.2.4.2 Seasonal attendance patterns
The seasonal change in attendance was inferred using Generalised
Additive Models
(GAMs) with a Poisson error structure in an exploratory fashion
(Hastie and Tibshirani
1990, Regular et al. 2010). GAMs were chosen to extend the
Generalised Linear Model by
application of smoothing terms to explanatory variables (Hastie
and Tibshirani 1990,
Wood 2006). Smoothing is undertaken by fitting splines, using
piecewise polynomials. The
degree of smoothing is defined by the number of knots supplied
(Fewster et al. 2000).
Selection of the number of knots is however not based upon an
objective criterion.
The number of knots can be determined through generalised cross
validation (GCV)
(Wood 2006, Forcada et al. 2006); however the number of knots
chosen must also consider
the scientific relevance of any observed pattern over time
(Fewster et al. 2000, Ruppert et
al. 2003).
By varying the number of knots the change in the pattern of
attendance is observed.
Final selection of the number of knots was based upon the point
at which the pattern of
attendance becomes poorly defined, with regards to the objective
(Fewster et al. 2000).
Whilst this approach is subjective displaying the results of a
range of knots demonstrates
the potential variation in defining change in attendance over
time and between colonies.
The robustness of any change in attendance over time is improved
by presenting
results of two different summaries of the response variable. The
form of the response
variable included using: 1) raw counts of the number of puffins,
and 2) taking the mean of
the number of puffins in each of the five ~5-hour periods each
day. Aggregating data into
five ~5-hour periods also reduced the proportion of zeros from
~60 % to ~20 %, improving
model fit. Corroboration between patterns of attendance from
GAMs using different
numbers of knots and the two response variables indicates the
persistence and robustness
of any pattern within the data.
-
Chapter 2 Attendance patterns of the Atlantic puffin
34
The degree of spatial synchrony between colonies is analysed
using the same
procedure as used for the correlation between weather variables
at two different sites. In
summary, data are de-trended by first-differencing and then,
given serial correlation, the
mean Pearson’s correlation is calculated by bootstrapping with
replacement (Bjornstad et
al. 1999, Buonaccorsi et al. 2001). Bootstrapping produces 95 %
confidence intervals to
also test the significance of any correlation when comparing
between time series.
Using the mean Pearson’s correlation from bootstrapping, all
comparisons showed
a positive and significant correlation between raw counts and
aggregated counts of puffins
(r > 0.95, 95 % confidence interval ranging from 0.92 -
1.00), for all colonies and with
different numbers of knots. Aggregating raw data into ~5-hour
periods and the inclusion of
daily period as a covariate further takes into consideration
autocorrelation within the model
residuals (Legendre and Legendre 2012). Therefore all further
analyses are based upon the
aggregated data. The seasonal attendance of puffins was not
modelled for Fair Isle given
extensive periods of missing data and thus the short length of
the time series (Table 2.1).
GAMs were run using the function gam from the ‘mgcv’ package
(Wood 2006, R.Core-
Team 2014).
For analysis of spatial synchrony between colonies on Mingulay
GAM-fitted values
using the aggregated data (from the five ~5 hour periods) were
correlated using the mean
Pearson’s correlation from bootstrapping.
2.2.4.3 Factors related to attendance
The effect of environmental variables (mean daily temperature
(˚C), mean daily wind
speed (km h-1
) and mean daily tidal height) on mean daily puffin attendance
was tested by
comparing the mean numbers of puffins before and during an
above-average
environmental event. An above-average event is defined as daily
temperature > 13 ˚C,
wind speed > 18.5 km h-1
(as defined within Hatch 2002), and the day of a spring or
neap
tide. These cut-off points for weather represent across colonies
the 70th
percentile for
temperature and the 58th
percentile for wind speed and were based upon a consideration
for
sample size and also prior study for wind speed (Hatch 2002).
Data were not analysed
using the entire daily mean time series given both the
confounding effect of seasonal
-
Chapter 2 Attendance patterns of the Atlantic puffin
35
attendance, which may mask any influence of environmental
variable on attendance and
the coarse scale of the covariate data.
The effect of temperature and wind speed compares the daily mean
number of
puffins between the days of an above-average event and the
preceding, consecutive day
with a value below the threshold. The effect of tide compares
the daily mean number of
puffins between the days of a spring or a neap tide and the
mid-point to the preceding neap
or spring tide, respectively. The midpoint prior to the
spring/neap peak or trough is ~3 days
before.
Comparisons of attendance before and during high temperatures,
winds or tidal
heights are made using a Wilcoxon matched-pairs signed-rank
test, given non-normality of
the count data. For weather variables data were analysed both
within individual colonies
and across colonies by grouping, using colonies on Mingulay and
from Hermaness. Tidal
data only from Mingulay were analysed by grouping across
colonies, given the small
sample size within a colony (n = 3). Hermaness was not analysed
given the presence of
only one tidal cycle (n = 1).
The effect of a predator’s presence on puffin numbers was
modelled using a
Wilcoxon matched-pairs signed-rank test, given non-normality of
the data. The data
comprised every instance of the paired numbers of puffins in two
consecutive photographs
without predators and then with predators, respectively. Given
ties within the analysis data
were jittered by adding a miniscule amount of random noise
(0.01) to the numbers of
puffins (Robert and Casella 2009). The Wilcoxon matched-pairs
signed-rank test was then
iteratively run 1000 times to model the significance of the
relationship with the addition of
random noise (Robert and Casella 2009).
Significance is reported as P < 0.05 with statistical tests
using a two-tailed test.
Error values associated with mean values are standard deviation
(SD), unless reported as
standard error (SE). Model selection to produce the minimum
adequate model (MAM) was
determined by stepwise backwards removal of parameters from a
fully parameterised
model using maximum likelihood (ML). The significance of a fixed
effect was determined
by using likelihood ratio (LRT) chi-square tests between the
GLMM and a GLMM
excluding the fixed effect of interest (Pinheiro and Bates 2000,
Miles 2010). Only
significant terms are reported from model selection. MAMs were
run using restricted
-
Chapter 2 Attendance patterns of the Atlantic puffin
36
maximum likelihood (REML) for presentation of model
coefficients. To test for
differences between model groups post-hoc Tukey Honest
Significant Difference tests are
used. The function lmer from the ‘lme4’ package was used to
implement GLMMs
(R.Core-Team 2014). Tests for autocorrelation were done using
the acf from the ‘stats’
package. Model fit was assessed on the basis of normalised model
residuals versus fitted
values testing for heteroskedascity, low AIC values, a lack of
over-dispersion and
collinearity. All analyses are carried out in R v. 3.03
(R.Core-Team 2014).
-
Chapter 2 Attendance patterns of the Atlantic puffin
37
Hermaness
Fair Isle
Mingulay
Figure 2.1. Location of the three islands where cameras were
installed. Top left: UK overview of island
locations. Camera positions denoted by triangles and monitored
puffin colonies in shaded ellipses.
-
Chapter 2 Attendance patterns of the Atlantic puffin
38
Results 2.3
2.3.1 Daily attendance patterns
Figure 2.3 shows the variation between colonies in daily
attendance from across the study
duration using counts of puffins aggregated into both 30 min
intervals and the five daily
~5-hour periods. There is a strong similarity in attendance
pattern between the two time
scales, and therefore the aggregated data from the five daily
~5-hour periods were analysed
to test for differences in mean attendance between colonies and
within a day. The results
showed a significant difference in daily attendance pattern
between colonies (LRT, period
of day*colony: χ2
24 = 221.68, P < 0.001). Within individual colonies post-hoc
Tukey tests
(as denoted by letters in Figure 2.3) showed on all Mingulay
colonies (except Otter 2012)
that mean attendance was greatest from the period termed Night
(19:12 to 00:00 h) to all
other periods of the day (Figure 2.3). There was also a
significant increase in attendance
from Early Morning (00:00 to 04:48 h) and Morning to Evening
(14:24 to 19:12 h) for
Original 2011 (Figure 2.3). On Hermaness mean attendance
significantly decreased from
Early Morning to all other periods of the day, except Night.
Mean attendance also
significantly increased from both Morning and Afternoon (09:36
to 14:24 h) to Night. Fair
Isle (both years) and Otter 2012 had no significant change in
puffin attendance across the
daily periods.
2.3.1.1 Pre- and post-hatching attendance
There are broadly similar daily attendance patterns before and
after hatching (Figure 2.4).
Within each breeding stage there is again similarity in the
daily attendance pattern of
puffins between the two aggregated time scales (Figure 2.4). The
average change in
attendance every 30 min across colonies on Mingulay between
breeding stages is shown in
Figure 2.5. Attendance appears to vary with breeding stage and
time of day, with a slight
increase in attendance from pre- to post hatching in the morning
before 07:00 h, and more
noticeably from 20:00 h onwards (Figure 2.5). This difference
across hatching stage
indicates a possible interaction between time of day and
breeding stage.
Using the five ~5 h daily periods showed that the effect of
breeding stage on
attendance depended on both colony and period of day for the
colonies on Mingulay (LRT:
-
Chapter 2 Attendance patterns of the Atlantic puffin
39
breeding stage*colony*period, χ2
24 = 79.47, P < 0.001). Model coefficients estimated that
there was significantly greater attendance from pre- to
post-hatching across colonies
(GLMM: Δpre- to post-hatching = 9.6 puffins ± 3.1 SE, P = 0.002,
Table 2.2). Post-hoc Tukey
tests showed that attendance increased significantly from pre-
to post-hatching for the
period Night for the colony Original 2012; whilst all other
colonies had a non-significant
increase for the same time through the breeding season. For all
other times of day there
were non-significant decreases in attendance with hatching, with
the exceptions of
Arnamul in the Early Morning and Evening, and Original 2011 in
the Evening, which both
showed a non-significant increase (Figure 2.4).
To examine the difference in attendance within different
breeding stages separate
GLMMs were run on each breeding stage (Table 2.3, Table 2.4).
During the pre-hatching
stage there was no significant interaction between period of day
and colony on the mean
puffin attendance, although an interaction existed in the
post-hatching stage (LRT:
colony*period; pre-hatching, χ2
12 = 18.72, P = 0.095; post-hatch, χ2
12 = 60.61, P < 0.001).
During the pre-hatching stage puffin attendance varied with both
period of day and colony
(LRT: period of day, χ2
4 = 20.45, P < 0.001; colony, χ2
3 = 30.37, P < 0.001). Post-hoc
Tukey tests showed that Arnamul had significantly higher numbers
of puffins than the
other colonies, whilst there were no significant differences in
attendance between the
colonies of Original 2011 and 2012, and Otter. Post-hoc Tukey
tests further showed
significant increases in mean numbers of puffins from the Early
Morning to Evening and a
significant increase from Early Morning to Night. During the
post-hatching stage post-hoc
Tukey tests showed a significant increase in mean numbers of
puffins throughout the day
and attendance peaked at Night, except for Otter, which had no
significant difference
between periods of day. Original 2011 also had significant
increases in attendance from the
two periods Early Morning and Morning to Evening (Figure
2.4).
2.3.2 Seasonal attendance pattern
Across the season there is considerable variation in attendance
for each colony from the
raw counts of puffins (Figure 2.6). Although, cyclicity at other
c